Table 3:

Multivariate regression models examining headache days/month, by cohort (late start vs. early start), adjusting for covariates.

Variables in the model:	Mean (95% CI) headache days/month in the late start vs. early start groups	Mean difference (95% CI) between groups in headache days/month
Unadjusted headache frequency	4.8 (4.3 to 5.3) vs. 7.7 (7.2 to 8.1)	−2.9 (−2.2 to −3.6)
Model #1: Adjusted for 1) total hours of sleep on school nights 2) sleep midpoint variability, 3) frequency of acute medication use, 4) being on a migraine preventive, 5) sex, 6) homework hours, 7) missing breakfast and 8) grade level	5.8 (5.3 to 6.2) vs. 7.1 (6.7 to 7.4)	−1.3 (−1.9 to −0.7)
Model #2: Adjusted for 1) getting at least 8 hours of sleep on school nights, 2) sleep midpoint variability, 3) frequency of acute medication use, 4) being on a migraine preventive, 5) sex, 6) homework hours, 7) missing breakfast, and 8) grade level	5.6 (5.2 to 6.0) vs. 7.1 (6.7 to 7.5)	−1.5 (−2.0 to −0.9)
Model #3: Adjusted for 1) total hours sleep on school nights, 2) sleep midpoint variability, 3) being on a migraine preventive, 4) sex, 5) homework hours, 6) breakfast missing, and 7) grade level	5.7 (5.2 to 6.2) vs. 7.7 (7.2 to 8.2)	−2.0 (−2.7 to −1.3)
Model #4: 1) Total hours sleep on school nights, 2) being on a migraine preventive, 3) sex, 4) homework hours, 5) breakfast missing, and 6) grade level	5.7 (5.2 to 6.2) vs. 7.7 (7.2 to 8.2)	−2.0 (−2.7 to −1.3)
Model #5: Adjusted for 1) sleep midpoint variability, 2) being on a migraine preventive, 3) sex, 4) homework hours, 5) breakfast missing, and 6) grade level	5.5 (5.0 to 6.0) vs. 7.8 (7.4 to 8.3)	−2.4 (−3.0 to −1.7)